<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->
wav2vec2-xlsr-ft-btb
This model is a fine-tuned version of techiaith/wav2vec2-xlsr-ft-cy on the TECHIAITH/BANC-TRAWSGRIFIADAU-BANGOR - NA dataset. It achieves the following results on the evaluation set:
- Loss: 0.3862
- Wer: 0.2817
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 0.0003
- train_batch_size: 16
- eval_batch_size: 8
- seed: 42
- gradient_accumulation_steps: 2
- total_train_batch_size: 32
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: linear
- lr_scheduler_warmup_steps: 500
- num_epochs: 5.0
- mixed_precision_training: Native AMP
Training results
Training Loss | Epoch | Step | Validation Loss | Wer |
---|---|---|---|---|
No log | 0.21 | 100 | 3.1084 | 1.0 |
No log | 0.41 | 200 | 1.4742 | 0.7157 |
No log | 0.62 | 300 | 0.7243 | 0.4482 |
No log | 0.83 | 400 | 0.5905 | 0.4043 |
2.399 | 1.03 | 500 | 0.5406 | 0.3665 |
2.399 | 1.24 | 600 | 0.5380 | 0.3699 |
2.399 | 1.44 | 700 | 0.4905 | 0.3580 |
2.399 | 1.65 | 800 | 0.5045 | 0.3727 |
2.399 | 1.86 | 900 | 0.4783 | 0.3474 |
0.6123 | 2.06 | 1000 | 0.4372 | 0.3272 |
0.6123 | 2.27 | 1100 | 0.4305 | 0.3225 |
0.6123 | 2.48 | 1200 | 0.4292 | 0.3141 |
0.6123 | 2.68 | 1300 | 0.4136 | 0.3233 |
0.6123 | 2.89 | 1400 | 0.4053 | 0.3140 |
0.4645 | 3.1 | 1500 | 0.4207 | 0.3261 |
0.4645 | 3.3 | 1600 | 0.4133 | 0.3130 |
0.4645 | 3.51 | 1700 | 0.4283 | 0.3118 |
0.4645 | 3.72 | 1800 | 0.4030 | 0.2982 |
0.4645 | 3.92 | 1900 | 0.3833 | 0.2911 |
0.3708 | 4.13 | 2000 | 0.3933 | 0.2966 |
0.3708 | 4.33 | 2100 | 0.3947 | 0.2902 |
0.3708 | 4.54 | 2200 | 0.3878 | 0.2849 |
0.3708 | 4.75 | 2300 | 0.3915 | 0.2854 |
0.3708 | 4.95 | 2400 | 0.3865 | 0.2811 |
Framework versions
- Transformers 4.28.1
- Pytorch 2.0.0+cu117
- Datasets 2.11.0
- Tokenizers 0.13.3